CAGEcleaner 1.0.0

Redundancy removal tool for cblaster hit sets

cagecleaner

Outline

cagecleaner removes genomic redundancy from gene cluster hit sets identified by cblaster. The redundancy in target databases used by cblaster often propagates into the result set, requiring extensive manual curation before downstream analyses and visualisation can be carried out.

Given a session file from a cblaster run (or from a CAGECAT run), cagecleaner retrieves all hit-associated genome assemblies, groups these into assembly clusters by ANI and identifies a representative assembly for each assembly cluster using skDER. In addition, cagecleaner can reinclude hits that are different at the gene cluster level despite the genomic redundancy, and this by different gene cluster content and/or by outlier cblaster scores. Finally, cagecleaner returns a filtered cblaster session file as well as a list of retained gene cluster IDs for easier downstream analysis.

Output

This tool will produce seven final output files - filtered_session.json: a filtered cblaster session file - filtered_binary.txt: a cblaster binary presence/absence table, containing only the retained hits. - filtered_summary.txt: a cblaster summary file, containing only the retained hits. - clusters.txt: the corresponding cluster IDs from the cblaster summary file for each retained hit. - genome_cluster_sizes.txt: the number of genomes in a dereplication genome cluster, referred to by the dereplication representative genome. - genome_cluster_status.txt: a table with scaffold IDs, their representative genome assembly and their dereplication status. - scaffold_assembly_pairs.txt: a table with scaffold IDs and the IDs of the genome assemblies of which they are part.

There are four possible dereplication statuses: - 'dereplication_representative': this scaffold is part of the genome assembly that has been selected as the representative of a genome cluster. - 'readded_by_content': this scaffold has been kept as it contains a hit that is different in content from the one of the dereplication representative. - 'readded_by_score': this scaffold has been kept as it contains a hit that has an outlier cblaster score. - 'redundant': this scaffold has not been retained and is therefore removed from the final output.

Installation

First set up a conda environment using the env.yml file in this repo, and activate the environment.

conda env create -f env.yml
conda activate cagecleaner

Then install cagecleaner inside this environment using pip. First check you have the right pip using which pip, which should point to the pip instance inside the cagecleaner environment.

pip install cagecleaner

Dependencies

cagecleaner has been developed on Python 3.10. All external dependencies listed below are managed by the conda environment, except for the NCBI EDirect utilities, which can be installed as outlined here.

• NCBI EDirect utilities (>= v21.6)
• NCBI Datasets CLI (v16.39.0)
• skDER (v1.2.8)
• pandas (v2.2.3)
• scipy (v1.14.1)
• BioPython (v1.84)
• more-itertools (v10.5)

Usage

cagecleaner expects as inputs at least the cblaster binary and summary files containing NCBI Nucleotide accession IDs. A dereplication run using the default settings can be started as simply as:

cagecleaner -b binary.txt -s summary.txt

Help message:

usage: cagecleaner [-c CORES] [-h] [-v] [-o OUTPUT_DIR] [-b BINARY_FILE] [-s SUMMARY_FILE] [--validate-files]
                  [--keep-downloads] [--keep-dereplication] [--keep-intermediate]
                  [--download-batch DOWNLOAD_BATCH] [-a ANI] [--no-content-revisit] [--no-score-revisit]
                  [--min-z-score ZSCORE_OUTLIER_THRESHOLD] [--min-score-diff MINIMAL_SCORE_DIFFERENCE]

   cagecleaner: A tool to remove redundancy from cblaster hits.
   
   cagecleaner reduces redundancy in cblaster hit sets by dereplicating the genomes containing the hits. 
   It can also recover hits that would have been omitted by this dereplication if they have a different gene cluster content
   or an outlier cblaster score.
   
   cagecleaner first retrieves the assembly accession IDs of each cblaster hit via NCBI Entrez-Direct utilities, 
   then downloads these assemblies using NCBI Datasets CLI, and then dereplicates these assemblies using skDER.
   If requested, cblaster hits that have an alternative gene cluster content or an outlier cblaster score 
   (calculated via z-scores) are recovered.
                                    

General:
 -c CORES, --cores CORES
                       Number of cores to use (default: 1)
 -h, --help            Show this help message and exit
 -v, --version         show program's version number and exit

Input / Output:
 -o OUTPUT_DIR, --output OUTPUT_DIR
                       Output directory (default: current working directory)
 -b BINARY_FILE, --binary BINARY_FILE
                       Path to cblaster binary file
 -s SUMMARY_FILE, --summary SUMMARY_FILE
                       Path to cblaster summary file
 --validate-files      Validate cblaster input files
 --keep-downloads      Keep downloaded genomes
 --keep-dereplication  Keep skDER output
 --keep-intermediate   Keep all intermediate data. This overrules other keep flags.

Download:
 --download-batch DOWNLOAD_BATCH
                       Number of genomes to download in one batch (default: 300)

Dereplication:
 -a ANI, --ani ANI     ANI dereplication threshold (default: 99.0)

Hit recovery:
 --no-content-revisit  Do not recover hits by cluster content
 --no-score-revisit    Do not recover hits by outlier scores
 --min-z-score ZSCORE_OUTLIER_THRESHOLD
                       z-score threshold to consider hits outliers (default: 2.0)
 --min-score-diff MINIMAL_SCORE_DIFFERENCE
                       minimum cblaster score difference between hits to be considered different. Discards outlier
                       hits with a score difference below this threshold. (default: 0.1)

   Lucas De Vrieze, 2025
   (c) Masschelein lab, VIB

Example case

We provide two example cases in the folder examples in this repo. We have already provided the cblaster output files as well as the original query fasta.

In the first case, 1146 gene cluster hits from Staphylococcus spp. should be reduced to 22 non-redundant hits. Running the cagecleaner for this example is done like below

cd N398V589S066P61
cagecleaner -s session.json -o output -c 20

This should give the seven output files in a new subfolder output. This should take about 10' using 20 cores, depending on the download speed of your internet connection. This requires 1.2 GB of disk space and 1.7 GB of RAM.

$ dir -1 output
clusters.txt
filtered_binary.txt
filtered_session.txt
filtered_summary.txt
genome_cluster_sizes.txt
genome_cluster_status.txt
scaffold_assembly_pairs.txt

In the second case, we queried four genes from MIBiG entry BGC0000194 (actinorhodin from Streptomyces coelicolor A3(2)), which yielded 8934 gene cluster hits. cagecleaner should reduce this to 4847 hits in about 1.5 h using 20 cores. 28.5 GB of disk space and 27.6 GB of RAM are required for this example case.

cd actinorhodin
cagecleaner -s session.json -o output -c 20

Citations

cagecleaner relies heavily on the skDER genome dereplication tool and its main dependendy skani, so we give these tools proper credit.

Salamzade, R., & Kalan, L. R. (2023). skDER: microbial genome dereplication approaches for comparative and metagenomic applications. https://doi.org/10.1101/2023.09.27.559801
Shaw, J., & Yu, Y. W. (2023). Fast and robust metagenomic sequence comparison through sparse chaining with skani. Nature Methods, 20(11), 1661–1665. https://doi.org/10.1038/s41592-023-02018-3

Please cite the cagecleaner manuscript:

In preparation

License

cagecleaner is freely available under an MIT license.

Use of the third-party software, libraries or code referred to in the References section above may be governed by separate terms and conditions or license provisions. Your use of the third-party software, libraries or code is subject to any such terms and you should check that you can comply with any applicable restrictions or terms and conditions before use.